ln-021-codegraph

Paths: File paths are relative to skills repo root.

Code Knowledge Graph

Type: Standalone Utility Category: 0XX Dev Environment

Indexes codebase into a layered graph (tree-sitter AST → SQLite) and provides dependency analysis, path tracing, references, implementations, and architecture overview via MCP tools.

Inputs

| Input | Required | Source | Description | |-------|----------|--------|-------------| | project_path | yes | args or CWD | Project root to index | | command | no | args | Specific action: index, search, symbol, paths, refs, arch |

When to Use

• Starting work on an unfamiliar codebase → index + architecture
• Before refactoring a function/class → find_symbols + inspect_symbol + trace_paths
• Understanding call flow → trace_paths
• Finding a symbol quickly → search

MCP Availability

Tool policy: follow host AGENTS.md MCP preferences; load references/mcp_tool_preferences.md and references/mcp_integration_patterns.md only when host policy is absent or MCP behavior is unclear.

Use hex-graph first when the task depends on symbol identity, references, implementations, architecture, dataflow, or edit impact. Use hex-line first for targeted local code reads when available. If MCP is unavailable, unsupported, or not indexed, continue with built-in Glob/Grep/Read/Bash, answer with manual evidence, and explicitly note the degraded confidence instead of blocking the skill.

Workflow

Phase 1: Index

Check if graph exists (.hex-skills/codegraph/index.db in project root).

If NOT exists:

Call: index_project({ path: "{project_path}" })

If exists (re-index on demand):

Call: index_project({ path: "{project_path}" })

Idempotent — skips unchanged files automatically.

If hex-graph is unavailable: build a manual project map with Glob and targeted Grep/Read, then continue to the closest matching query workflow without indexing.

Phase 2: Query

Route based on user intent:

| User says | Tool | Parameters | |---|---|---| | "Show dependencies" / "What uses X?" | trace_paths | { name: "X", file: "...", path_kind: "mixed", direction: "reverse", path: "{project_path}" } | | "Who calls X?" / "What does X call?" | trace_paths | { name: "X", file: "...", path_kind: "calls", direction: "reverse"\|"forward", path: "{project_path}" } | | "Tell me about X" / "Context of X" | inspect_symbol | { name: "X", file: "...", path: "{project_path}" } | | "Project structure" / "Architecture" | analyze_architecture | { path: "{project_path}", scope?: "src/" } | | "Find symbol X" | find_symbols | { path: "{project_path}", query: "X" } | | "Find app.get(...) / router.use(...) / server.registerTool(...) pattern" | grep_search | { path: "{project_path}", pattern: "app\\.get\\(|router\\.use\\(|server\\.registerTool\\(" } | | "Find duplicate code / hotspots / unused exports" | audit_workspace | { path: "{project_path}", scope?: "src/", verbosity: "minimal", limit: 5, clone_member_limit: 3 } | | "Circular dependencies / module coupling" | analyze_architecture | { path: "{project_path}", verbosity: "full" } | | "Implementations / overrides" | find_implementations | { name: "X", file: "...", path: "{project_path}" } | | "Dataflow / propagation" | trace_dataflow | { source: { symbol: { name: "X", file: "..." }, anchor: { kind: "param", name: "input" } }, sink?: { symbol: { name: "X", file: "..." }, anchor: { kind: "return" } }, path: "{project_path}" } | | "Review a diff / worktree" | analyze_changes | { path: "{project_path}", base_ref: "origin/main" } | | "Check what editing this range affects" | analyze_edit_region | { path: "{project_path}", file: "src/file.ts", line_start: 10, line_end: 40 } |

Canonical selector rule: Semantic tools accept exactly one selector:

• symbol_id
• workspace_qualified_name
• qualified_name
• name + file

Preferred flow: use find_symbols only after narrowing path as much as practical, then feed the returned workspace_qualified_name into inspect_symbol, trace_paths, find_references, or find_implementations for exact follow-up queries.

Query boundary rule: find_symbols is name-based discovery only. For code fragments like export function or unresolved member-call patterns like app.get(...), use grep_search instead of treating them as symbols.

Ambiguity rule: if find_symbols returns partial ... truncated=1 or shows more total results than returned rows, refine with path, then name + file or workspace_qualified_name instead of widening the graph query.

Path rule: path may be the indexed project root or any file/subdirectory inside that indexed project.

Dataflow anchors: trace_dataflow requires source.anchor and optional sink.anchor. Use:

• param for function parameters
• local for local variables
• return for function returns
• property with access_path for bounded property flow

Precision controls: inspect_symbol, trace_paths, and find_references support min_confidence (low, inferred, exact, precise) when the caller wants to suppress weaker parser-only facts.

Phase 3: Present Results

1. Show MCP tool output directly when available; otherwise present manual findings and mark them as fallback reasoning
2. For code snippets referenced in results, use hex-line read_file with line ranges when available; otherwise use built-in Read
3. Suggest follow-up queries based on results:
4. After find_symbols with a clean top match → suggest inspect_symbol with workspace_qualified_name
5. After find_symbols with partial ... truncated=1 → suggest narrowing path or switching to name + file before any deeper graph tool
6. After inspect_symbol → suggest trace_paths if refactoring
7. After trace_paths → suggest find_references or find_implementations depending on symbol kind
8. After empty trace_paths from a broad or module-level selector → suggest inspect_symbol or analyze_architecture instead of assuming there are no dependencies

Supported Languages

| Language | Extensions | Coverage | |---|---|---| | JavaScript | .js, .mjs, .cjs, .jsx | Strongest semantic coverage | | TypeScript / TSX | .ts, .tsx | Strongest semantic coverage | | Python | .py | Workspace-aware definitions, calls, imports, unused exports; optional precise overlay when provider is installed | | C# | .cs | Workspace-aware definitions, calls, project/namespace ownership, type relations; optional precise overlay when provider is installed | | PHP | .php | Workspace-aware definitions, calls, PSR-4 namespace imports, unused exports; optional precise overlay when provider is installed |

MCP Server Setup

Add to .mcp.json:

{
  "mcpServers": {
    "hex-graph": {
      "command": "node",
      "args": ["{skills_repo}/mcp/hex-graph-mcp/server.mjs"]
    }
  }
}

Definition of Done

• [ ] Project indexed or manual fallback map built
• [ ] Query results shown to user
• [ ] Follow-up suggestions provided
• [ ] Fallback path stated when MCP was unavailable

Version: 0.1.0 Last Updated: 2026-03-20